Image processor for character recognition
In image processing of a document having image data of a document including an image with character images on a background image, the character images and the background image thereof are separated. First, areas in correspondence to the character images from the image data are extracted, and character code data are generated based on the extracted areas in the image data. On the other hand, the character image in the image data is replaced with the background image. The character images are reproduced with reference to the character code data at the positions of the character images. On the other hand, the original character images are deleted from the image data. The character images on the background image is complemented based on the background image data. Then the character images based on the character code data and the background image thereof are synthesized.
Latest Minolta Co., Ltd. Patents:
- Communication system and method in public key infrastructure
- Fluid transferring system and micropump suitable therefor
- Network device connecting system, data transmission device, data receiving device, and portable terminal
- Network device connecting system, data transmission device, data receiving device, and portable terminal
- Method and apparatus for improving edge sharpness with error diffusion
This application is a Divisional Application of application Ser. No. 09/875,081, filed Jun. 7, 2001, which in turn is based on application Ser. No. 2000-173727 filed in Japan, the contents of which are hereby incorporated by reference.
BACKGROUND OF THE INVENTION1. Field of the Invention
The present invention relates to optical character recognition which converts character images in image data to character code data.
2. Description of Prior Art
In character recognition by using an optical character recognition apparatus, characters in a text image in a document are recognized. As to a document including text image and the other image, it is known to further recognize the layout of the document and to arrange data of the other image at relevant positions in the layout as bit map image data. The layout, the character code data and the bit map image data are outputted as one file.
However, when a document has characters on a background image, after converting the character image data to character code data, it is difficult to synthesize the recognized characters (character images based on, the character code data) with the original image data. This is due to difference in the font and the positions of the characters in the original image data with the counterparts based on the conversion data.
In prior art character recognition, it is known, for example, as shown in
An object of the present invention is to provide image processing which can reproduce characters on the background image well.
In the present invention, character images and the background image thereof are separated in image: data of a document including an image with the character images on the background image. First, areas in correspondence to the character images from the image data are extracted, and character code data are generated based on the extracted areas in the image data. On the other hand, the character image in the image data is replaced with the background image. The character images are reproduced with reference to the character code data at the positions of the character images. On the other hand, the original character images are deleted from the image data. The character images on the background image is preferably complemented based on the background image data. Then the character images based on the character code data and the background image thereof are synthesized. Thus, the synthesized image is reproduced well.
An advantage of the present invention is that a document image is reproduced well for character images on the background image.
BRIEF DESCRIPTION OF THE DRAWINGSThese and other objects and features of the present invention will become clear from the following description taken in conjunction with the preferred embodiments thereof with reference to the accompanying drawings, and in which:
Referring now to the drawings, wherein like reference characters designate like or corresponding parts throughout the several views,
The computer 200 has software programs such as a scanner driver 230 and a printer driver 240 which control the peripheral devices. For example, the scanner driver 230 instructs to set reading conditions and to start reading for the scanner 400, in a screen in the computer 200.
In a document recognition processor 210 in the computer 200, images on the background image in a document are separated into the character images and the background bit map image. Characters are recognized in the character images with use of optical character recognition technique, while the character images are deleted from the image data. Then, they are synthesized again to reproduce the original document.
In this embodiment, the computer 200 has the function of the character recognition, but the character recognition function may also be incorporated in an image input apparatus or an image output apparatus. The image output apparatus is not limited to the printer 300, and it may be, for example, a display device. The image input apparatus is not limited to the scanner 400, and it may be a device such as a digital camera or a film scanner which inputs a digital image.
In the document recognition processor 210, a document recognition processor 210 extracts character image data from image data which may include a character image on a background image and converts the extracted character image data to character code data with optical character recognition. Further, the image data of the character image on a background image is complemented with reference to the ambient background image data. Then, image data based on the character code data are synthesized with the complemented image data. Practically, a character recognizer 212 recognizes character image data in the image data obtained by reading a document and converts the character image data to character code data. A character deleter 214 deletes the recognized character code data from the original image data, and an image recognizer 216 recognizes image data such as a photograph or a design other than the white space (or an area having no image). An image synthesizer 218 synthesizes the character code data with the image data recognized by the image recognizer 216. Thus, the character image data included in the original document does not overlap the character code data. Further, by storing the character code data and the background image included in a document in the storage device 205, the character image data included in the document can be edited or changed by using the character code data. The character code data can also be used as key words for search.
Next, preprocessing for optical character recognition (OCR) is performed on the image data in the converted color space, in order to enhance recognition efficiency in the character recognizer 212 (S12). Practically, the image data of L, a and b components are subjected to binarization so that character image data become black and the background image becomes white. The image data of L, a and b components are stored in a different place (memory). Further, in the preprocessing of optical character recognition, for example, noises included in the image data such as an isolated point may be deleted, inclination of the image data due to misplacement of a document on document reading may be corrected, or deformed characters may be corrected. In this example, the preprocessing is performed on the image data of L, a and b components, but it may be performed on the image data of R, G and B components before the conversion of color space.
Next, in the character recognizer 212, lines consisting of character image data included in the image data and each character image data in the lines are taken out from the image data. Then, character code, font, font size and the like are recognized on each character image data (S14). The character code data is determined, for example, for each character in this embodiment. It may also be performed in the unit of word. As to the font and font size, the most similar font and font size are selected among the font data group in the document recognition processor 210. The character recognizer 212 outputs position coordinates (X1, Y1) and (X2, Y2) of a rectangle having the origin (0, 0) at the top leftmost point, as shown in
When color is changed within a character, the character is regarded a result of erroneous recognition on an image other than the character, and it is not converted to character code data. That is, character image data including color change is not converted to character code data. In the preprocessing for optical character recognition (S12), the binarization is performed on the image data of L, a and b components, while the image data is stored in a different place. The color of a character is detected from the above-mentioned stored image data. Practically, by using the 3*3 maximum filter as shown in
wherein N denotes a number of pixels of the j-th character.
In the decision of color change, the circumscribing rectangle of a character after the subtraction is divided into four areas as shown in
Cj1=(Lj1, aj1, bj1),
Cj2=(Lj2, aj2, bj2),
Cj3=(LJ3, aj3, bj3), (2)
and
Cj4=(Lj4, aj4, bj4).
It is decided that the color is changed when the following condition is satisfied,
|Ljn-Llm|≧2 k1,
|ajn-ajm|≧k2, (3)
or
|bjn-bjm|>k3,
wherein n=1, 2, 3 or 4, and m=1, 2, 3 or 4, and k1, k2 and k3 denote constants. The above-mentioned average image data Cj of the character is determined actually by determining the averages of the four areas first and by further averaging the four averages. When the color is changed in a character, the above-mentioned conversion to character code data is not performed.
Next, in the character deleter 214, character image data recognized in the character recognizer 212 is deleted from the original image data of L, a and b components (S16). As mentioned above, the original image data are stored in the different place in the binarization process in the preprocessing at step S12. In the character deletion, the binarized image data are subjected to filtering with use of a 5*5 minimum filter shown in
Next, the image data of L, a and b components after the conversion to white is subjected to filtering with a 7*7 filter shown in
An object pixel shown in
The bit map image data only of the background image generated as mentioned above and the character code data recognized in the character recognizer 212 are arranged and synthesized as shown in
In the above-mentioned embodiment, characters on the background image in a document are separated into the character code data and the background bit map image, from which characters are deleted, and they are synthesized again to reproduce the document. Therefore, the system has following advantages.
(1) The encoded characters are not reproduced doubly with the character image, and the output image reproduces the background of the document.
(2) Further, because characters on the background can be encoded, the output character image can be read easily, the character size can be changed, or the characters can be used for search.
(3) Still further, data in the background can be used again. For example, when the characters are modified, a document having the same background can be generated.
(4) Because character image data having a changing color is not converted to character code data, erroneous recognition of non-characters as characters can be decreased.
Although the present invention has been fully described in connection with the preferred embodiments thereof with reference to the accompanying drawings, it is to be noted that various changes and modifications are apparent to those skilled in the art. Such changes and modifications are to be understood as included within the scope of the present invention as defined by the appended claims unless they depart therefrom.
Claims
1. A method of image processing comprising the steps of:
- (a) obtaining image data of an original image, the original image having character images provided on a background image;
- (b) conducting character recognition of the image data as a first process, including the steps of: (b-1) preprocessing the image data for character recognition; (b-2) extracting areas from the image data which correspond to the character images; and (b-3) generating character code data based on the extracted character images;
- (c) conducting a second process on the image data based on the first process, including the step of replacing the extracted areas with the background image with reference to the image data; and
- (d) storing data obtained by conducting the second process on the image data and the character code data along with a relationship between them.
2. The method of claim 1, wherein in the step (a) the image data is obtained from an image reader which optically reads the original image on a document and generates the image data.
3. The method of claim 1, wherein in the step (b-3) the character code data are generated in reference to color information of the character images.
4. The method of claim 3, wherein in the step (b-3) the character code data are not generated from a character image that has color change.
5. The method of claim 1, wherein in the step (b-2) the areas are extracted character by character.
6. The method of claim 1, wherein in the step (b-2) the areas are extracted in the unit of word.
7. An image processing apparatus comprising:
- an image reader which reads image data of an original image, the original image having character images provided on a background image;
- character recognition device conducting character recognition of the image data including: a preprocessing device preprocessing the image data for character recognition; an extractor which extracts areas from the image data which correspond to the character images; and a converter which generates character code data based on the extracted character images;
- a replacer which replaces the extracted character images with the background image with reference to the image data; and
- a storage device which stores the replaced image data and the character code data along with a relationship between them.
8. The apparatus of claim 7, wherein the image reader optically reads the original image on a document and generates the image data.
9. The apparatus of claim 7, wherein the character code data are generated in reference to color information of the character images.
10. The apparatus of claim 9, wherein the character code data are not generated from a character image that has color change.
11. The apparatus of claim 7, wherein the extractor extracts character images character by character.
12. The apparatus of claim 7, wherein the extractor extracts character images in the unit of word.
13. A computer program product for image processing comprising computer-readable instructions for performing the steps of:
- (a) obtaining image data of an original image, the original image having character images provided on a background image;
- (b) conducting character recognition of the image data as a first process, including the steps of: (b-1) preprocessing the image data for character recognition; (b-2) extracting areas from the image data which correspond to the character images; and (b-3) generating character code data based on the extracted character images;
- (c) conducting a second process on the image data based on the first process, including the step of replacing the extracted areas with the background image with reference to the image data; and
- (d) storing data obtained by conducting the second process on the image data and the character code data along with a relationship between them.
14. The computer program product of claim 13, wherein in the step (a) the image data is obtained from an image reader which optically reads the original image on a document and generates the image data.
15. The computer program product of claim 13, wherein in the step (b-3) the character code data are generated in reference to color information of the character images.
16. The computer program product of claim 15, wherein in the step (b-3) the character code data are not generated from a character image that has color change.
17. The computer program product of claim 13, wherein in the step (b-2) the areas are extracted character by character.
18. The computer program product of claim 13, wherein in the step (b-2) the areas are extracted in the unit of word.
19. A computer readable medium encoded with computer executable instructions for image processing comprising the steps of:
- (a) obtaining image data of an original image, the original image having character images provided on a background image;
- (b) conducting character recognition of the image data as a first process, including the steps of: (b-1) preprocessing the image data for character recognition; (b-2) extracting areas from the image data which correspond to the character images; and (b-3) generating character code data based on the extracted character images;
- (c) conducting a second process on the image data based on the first process, including the step of replacing the extracted areas with the background image with reference to the image data; and
- (d) storing data obtained by conducting the second process on the image data and the character code data along with a relationship between them.
20. The computer readable medium of claim 19, wherein in the step (a) the image data is obtained from an image reader which optically reads the original image on a document and generates the image data.
21. The computer readable medium of claim 19, wherein in the step (b-3) the character code data are generated in reference to color information of the character images.
22. The computer readable medium of claim 21, wherein in the step (b-3) the character code data are not generated from a character image that has color change.
23. The computer readable medium of claim 19, wherein in the step (b-2) the areas are extracted character by character.
24. The computer readable medium of claim 19, wherein in the step (b-2) the areas are extracted in the unit of word.
Type: Application
Filed: Jun 1, 2007
Publication Date: Oct 11, 2007
Patent Grant number: 7454060
Applicant: Minolta Co., Ltd. (Osaka)
Inventor: Yoko Fujiwara (Tokyo)
Application Number: 11/806,642
International Classification: G06K 9/18 (20060101);